Dr. Carles Méndez-Ortega (p), Dr. Albert Miro Pérez (p), Dr. Joan Torrent-Sellens

This paper examines the heterogeneous effects of predictive artificial intelligence (AI) adoption on total factor productivity (TFP) among Spanish manufacturing firms, with particular emphasis on the role of the COVID-19 pandemic as a technological turning point. Predictive AI—encompassing advanced data analytics, cloud computing, and algorithm-based forecasting tools—has become increasingly central to firm-level decision-making and operational efficiency. Yet, empirical evidence on which firms benefit most from these technologies remains limited. Using firm-level data from the Encuesta sobre Estrategias Empresariales (ESEE) for 2018 and 2022, we analyze how the productivity effects of predictive AI vary across the TFP distribution and how these effects evolved before and after the pandemic. Our empirical strategy combines quantile regression techniques to capture productivity heterogeneity with two-stage least squares (2SLS) estimations to address potential endogeneity in AI adoption. The results reveal that the positive impact of predictive AI on TFP has become stronger and more widespread in the post-pandemic period, with particularly pronounced effects among firms located in the upper quantiles of the productivity distribution. Firm-level capabilities—such as training investment, international market engagement, and process innovation—consistently enhance productivity outcomes. In addition, regional capital intensity supports firm productivity, while strong localization economies exert negative spillover effects. Overall, the findings highlight the growing productivity returns to predictive AI adoption and underscore the importance of firm capabilities and regional context in shaping these gains.

Mr Fernando de la Torre Cuevas (p), Mr Kurt Kratena

Teleworking is arguably one of the most notable effects of the digital transition in developed countries. After the COVID-19 pandemic, it seems that telework will stick, at least to some extent. Admittedly, a number of studies have now assess the energy impacts of telework for alternative case studies. This paper contributes to the existing literature using a multiregional input-output (MRIO) model and relaxing some of the assumptions behind energy impact assessments of telework in two different ways. First, we couple the MRIO model with a job-residential location choice model in which preferences for local amenities are not fixed; rather, they are an increasing function of telework intensity. Second, regarding production and consumption, we relax some of the rigidities in Computable General Equilibrium models so that some Keynesian multiplier effects may appear. Furthermore, our model disaggregates household consumption patterns by (1) region degree of urbanisation. Each commodity has its own energy consumption coefficient per unit of output. We calibrate our model to replicate the Austrian multiregional economy in 2021. We then shock our model restricting telework and simulating an alternative set of job-residential location choices. Individual choices result in alternative aggregate household consumption and labour supply. The model is solved iteratively until reaching a new equilibrium, thus providing alternative regional energy consumption levels as well. We interpret the gap between the observed and alternative energy consumptions as the energy impacts of telework.

Dr. Albert Miró-pérez (p), Dr. Ángeles Pereira-Sánchez, Dr. Joan Torrent-Sellens

This paper examines the determinants of eco innovation and social innovation in Spanish firms, drawing on microdata from the 2022 Encuesta sobre Estrategias Empresariales (ESEE). While previous research has highlighted the relevance of technological capabilities and environmental regulation for eco innovation, empirical evidence on the joint drivers of environmental and social innovation and their organizational antecedents, remains limited. Building on the Technology–Organization–Environment (TOE) framework and the resource based view, this study estimates a set of Probit models to analyse how strategic sustainability orientation, environmental financial effort, technological capabilities, and organizational barriers shape firms’ propensity to innovate sustainably.
Results reveal a clear asymmetry between eco innovation and social innovation. Eco innovation is significantly influenced by specific organizational barriers, particularly coordination failures and resource needs, while social innovation appears less sensitive to internal obstacles. In both domains, environmental investment and environmental expenditure emerge as the most robust predictors of sustainable innovation, confirming strong complementarities between environmental commitment and broader sustainability outcomes. Technological capabilities, especially digitalization and R&D intensity, consistently increase the likelihood of innovating, whereas firms operating in high technology sectors exhibit a lower propensity to adopt sustainable innovations. Organizational barriers related to knowledge and innovation capabilities show negative effects, particularly in models with aggregated indices.
Incremental TOE models confirm that technological and environmental factors dominate the explanation of sustainable innovation, while organizational factors display heterogeneous and often inhibitory effects. Overall, the findings underscore the importance of strengthening firms’ technological capabilities and environmental investment efforts, while addressing internal organizational rigidities that hinder the transition toward more sustainable business models.

Prof. Roberta Capello, Dr. Carolina Foglia (p), Prof. Camilla Lenzi

The debate on European technological sovereignty has intensified in light of persistent productivity gaps, asymmetric dependencies, and geopolitical pressures. While current policy frameworks such as the Open Strategic Autonomy (OSA) agenda and the EU’s Economic Security strategy define sovereignty at the continental level, they often overlook the profound regional heterogeneity that characterizes Europe’s innovation systems. This paper proposes a regionalized and operational definition of technological sovereignty, grounded in innovation systems theory and the concept of sovereignty as embedded agency and strategic openness.
The objective of the contribution is to identify and measure three distinct modes of regional knowledge creation: local own capacity, interregional own capacity, and external collaboration, and assess their individual and combined effects on regional productivity. Methodologically, the paper builds an original matched dataset linking AI patents, firms, and regions, and computes sovereignty indicators based on patent ownership and transaction flows. The empirical strategy relies on population-weighted panel regressions (2015–2023) associating Total Factor Productivity (TFP) levels with lagged sovereignty indicators and their interaction, controlling for structural and geographical characteristics.
The originality of the work lies in its non-protectionist, non-autarkic operationalization of technological sovereignty, its focus on patent ownership and invention, and its ability to unpack sovereignty components at the regional level.

Prof. Andrés Rodríguez-Pose (p), Mr Federico Bartalucci

Artificial intelligence (AI) is widely expected to transform economies, yet evidence on its effects on regional economic growth remains limited. This paper distinguishes between two channels through which AI shapes economic outcomes: AI generation —the creation of AI innovations— and AI adoption —the deployment of AI technologies. Using panel data for European regions over 2013–2023, we measure AI generation, using AI patents as a proxy, and construct a latent measure of AI adoption anchored in observed firm-level uptake to compare their relationships with GDP per capita growth. Fixed-effects estimates reveal a clear asymmetry. AI adoption, which is geographically diffuse across European regions, is linked with significant gains in GDP per capita. AI generation, which is heavily concentrated in a handful of metropolitan regions, displays a weaker and largely insignificant relationships with local economic outcomes. Interaction effects show that AI adoption also enhances the returns to regional investments in R&D and higher education. The findings point towards AI strategies that prioritise adoption alongside investment in complementary capabilities, rather than focusing narrowly on indigenous AI generation. Such an approach would not only prevent greater economic polarisation within Europe but also make a more significant contribution to the overall competitiveness of the continent.